Rishi Jaluria

Peter described a simple predictable AWS-style price with usage overage charges, aiming for mostly predictable billing that large customers prefer.

Peter clarified Falcon is standard agentic labor to solve industry problems, not customer-facing AI tooling; custom agents use Vault AI.

Peter emphatically stated AI is not replacing core systems of record like SAP, Workday, or Veeva. LLM providers are an engine for the new wave of cloud computing, the new AWS. Relationship is symbiotic: Veeva couldn't build AI applications without foundational LLMs, just as Veeva couldn't exist without cloud infrastructure. AI will play out over 10-20 years.

Peter said 90% of value in life sciences is lost between diagnosis and correct treatment, not in drug discovery. AI doctors and LLMs can bridge the gap between patients, doctors, and pharma. Veeva will play a part connecting the industry to its external ecosystem. Benefits of AI over 10-20 years are underestimated.

Peter said migration tooling is well-refined non-AI tooling. Over 20 migrated with ~300 more to go. AI will create pull for migration though. Expects more than two-thirds of migrations done in 2026 and 2027.

Peter gave example of filing a document to TMF from Microsoft Office via agent-to-agent protocol. MCP enables much less brittle integration than traditional middleware. Cross-vendor agent interoperability is an underappreciated benefit.

Paul said bringing sales, marketing, and service together on a single customer database is resonating. Currently a smaller pool since customers must be on Vault CRM first, largely focusing on smaller and mid-sized customers.

Peter said AI is deeply embedded in core applications (MLR agent, CRM agent, safety AI agent). LLM technology has settled down, and Veeva is building nondeterministic computing into specific use cases.

Peter said the LLM landscape is stabilizing as expected. Veeva is now investing more in AI solutions with a centralized group and strong leader.

Leighton framed compute as a core-to-edge continuum (robotics example of video ingestion plus low-latency inference); Akamai serves inference and small/medium models, not frontier training.

Ed confirmed CIS is 100% additive to existing customer spend, not substituting for delivery revenue. The large deal was done out-of-cycle, not as part of a renewal. Good success with both existing and new customers. New verticals emerging that weren't traditional Akamai CDN customers. Total new customer count has picked up over the last 18 months.

Ed broke down event revenue impacts: Super Bowl/concerts are small ($0.5-1M capacity fees), Olympics a few million, World Cup $3-6M, NFL season much better. Gaming releases depend on popularity; console refresh cycles are larger (hundreds of millions of firmware updates). Events are nice to have but not overly material for a $1B+ quarter company.

Tom emphasized it is not about capacity constraints. One hyperscaler uses Akamai for ad logic proximity, another for API orchestration, a third for media workflow via managed container service. It is about Akamai's unique 4,000+ POP distributed platform delivering faster performance.

Tom confirmed they are already extending API security to detect AI agents, shadow AI apps, and protect inference applications with AI firewall. Sees this as one of the reasons they are so excited about API security business.

Tom emphasized avoiding account breakage by phasing changes over two years. Adding specialized sellers for new products rather than reassigning accounts all at once.

Ed acknowledged fewer players could help pricing but noted delivery is still volume-driven. Seeing some rationalization with longer contract terms.

Praveer said new products deepen seats in the base and accelerate lower-end customer acquisition, with agent broadening paid users.

Dylan kept focus on the code-to-canvas round-trip and model-provider integrations over formal partnerships.

Praveer described credit telemetry, co-termed add-on packs, and pay-as-you-go for burst usage as predictability levers.

Praveer said Make weekly usage among $100K+ customers keeps growing and increasingly anchors the sales narrative.

Dylan argued agents will multiply design surfaces and increase the designer's role rather than replace design work.

Khozema said AI-native customers are still early and relatively small, and the bigger opportunity is enterprise adoption, which is moving more slowly on buying cycles.

Khozema said a simpler self-serve experience serves international customers too, with boots on the ground selectively in Australia, Japan and Singapore.

Khozema declined to give a profile but cited cost discipline, workforce mix shift, AI and automation, and a buyback grid as ongoing levers.

Dev outlined why MongoDB fits agents, JSON to model messy real-world data, integrated search/vector, and memory to connect context across sessions; well positioned as agents deploy.

Dev said AI's accelerating pace of change makes rigid relational databases untenable, positioning MongoDB's flexible document model, search, and Voyage embeddings for net-new AI apps.

Satya explained seats as convenient consumption packs with overages flowing to pure consumption billing. Long-term commitments get appropriate discounting. Value is evaluated by evals and outcomes on token-level economics.

Satya cited fastest-ever M365 suite adoption; Amy explained Copilot impact shows in M365 commercial ARPU, currently masked by lower-ARPU frontline and SMB seat growth.